Computer Vision News - December 2016

24 Computer Vision News Application Application real social needs, beyond just social networking or search engines. During development, big data analysis proved difficult. They meet these challenges by taking advantage of deep learning techniques : in fact, while data is very noisy, deep learning provides a promising direction for this system. Meanwhile, they wanted to tackle the issue of personalization, especially when diagnosing patients who may exhibit varying symptoms to different degrees. Deep learning is an excellent technique to do that. From an algorithmic point of view, they want to propose a multi-modal fusion technique to the research community. The technique looks at how multi- modality can be associated and mapped with personalization in end-to- end assisted living for elderly people. In order to create a successful product that can work in real life, it’s almost impossible to have the label or ground truth for every situation . When they detect anomalies in a person’s behavior, the system cannot know what is normal for someone and what is not. Therefore, a supervised approach did not work very well in this case. Instead, unsupervised approach studies a person’s behavior without human intervention, using different clustering and classification technique to detect and track the behavior of people in a room, such as where people stand and sit or how fast they move. ICT4LIFE will make all their datasets and publications open to the public. They continue to work towards making the alpha version of their system available during their first pilot by June 2017 . There are also funny moments while working on the project. Stylianos, Dario and Mirela told us that, as a part of their research, they developed data sets of activity recognition in which they recorded people asked to find different objects hidden in a room. Sometimes they would ask people to find objects that weren’t there at all. Of course, the person’s reaction was always entertaining as they continued to search in all of the drawers and cupboards without any success. Here is Mirela, looking for something that is not.